Linking to Data¶

Need metadata and data

Linking to Data¶

  • linked-data metadata models describe structure + composition of buildings
  • this metadata provides context to better understand building data telemetry
  • how do we associate data with the metadata model?
    • it's another "link"!

Sources of Data and Telemetry¶

  • some common ones:
    • historical telemetry (timeseries data)
    • building geometry, BIM
    • building energy models (BEM)
    • live telemetry (OPC-UA, BACnet, Modbus)

Sources of Data and Telemetry¶

  • link to these other sources:
    • preserve existing investments in models + tools
    • avoid unnecessary duplication
  • often other DBs are better at those data anyway...

ref

Looking at an Example Model¶

# load in the graph
import brickschema
g = brickschema.Graph()
g.load_file("ttl/Brick.ttl")
g.load_file("ttl/building-compiled.ttl")
print(f"Graph has {len(g)} triples")

Graph has 53476 triples

# what kinds of entities are in the model?
df = sparql_to_df(g, """SELECT ?entity ?type WHERE {
  ?entity a ?type .
  FILTER(REGEX(STR(?entity), "^http://openmetrics.eu/openmetrics#"))
}""")
df.groupby('type').count().sort_values(by='entity', ascending=False).head(10)
entity
type
https://w3id.org/bot#Element 326
https://brickschema.org/schema/Brick#TimeseriesReference 323
https://pi.pauwel.be/voc/buildingelement#Wall 204
https://pi.pauwel.be/voc/buildingelement#Door 48
https://brickschema.org/schema/Brick#Space 36
https://w3id.org/bot#Space 36
https://brickschema.org/schema/Brick#Zone 19
https://brickschema.org/schema/Brick#Terminal_Unit 18
https://brickschema.org/schema/Brick#Building_Electrical_Meter 18
https://brickschema.org/schema/Brick#Thermostat 18 
# what kinds of data sources are in the model?
df = sparql_to_df(g, """SELECT ?entity ?type WHERE {
  ?entity rdf:type/rdfs:subClassOf* brick:Point .
  ?entity a ?type .
  FILTER(REGEX(STR(?entity), "^http://openmetrics.eu/openmetrics#"))
}""")
df.groupby('type').count().sort_values(by='entity', ascending=False).head(10)
entity
type
https://brickschema.org/schema/Brick#Active_Power_Sensor 17
https://brickschema.org/schema/Brick#Delay_Parameter 17
https://brickschema.org/schema/Brick#Max_Temperature_Setpoint_Limit 17
https://brickschema.org/schema/Brick#Min_Temperature_Setpoint_Limit 17
https://brickschema.org/schema/Brick#Mode_Command 17
https://brickschema.org/schema/Brick#Occupancy_Sensor 17
https://brickschema.org/schema/Brick#On_Off_Command 17
https://brickschema.org/schema/Brick#Speed_Setpoint 17
https://brickschema.org/schema/Brick#Zone_Air_Temperature_Sensor 17
https://brickschema.org/schema/Brick#Zone_Air_Temperature_Setpoint 17

Let's zoom into one of the "spaces" and see what is modeled there:

  • bold outlined nodes are other "entities" in the model
  • color-filled nodes are "classes" or "types"
  • edges represent properties and relationships
space_graph = get_local_graph(g, URIRef("http://openmetrics.eu/openmetrics#Space_787"))
display.display_png(viz(space_graph))

Use queries against the model to find interesting/relevant data sources, then go fetch the data and analyze!

Find Sensors in Spaces¶

Write quick SPARQL query to find sensors which measure locations

df = sparql_to_df(g, """SELECT ?sensor ?data_id ?location WHERE {
   ?sensor rdf:type/rdfs:subClassOf* brick:Sensor .
   ?sensor ref:hasExternalReference/ref:hasTimeseriesId ?data_id .
   ?sensor brick:isPointOf ?location .
   ?location rdf:type/rdfs:subClassOf* brick:Location
}""")
df.head()
sensor data_id location
0 http://openmetrics.eu/openmetrics#RC04N0061_Oc... TUC.245.77.R4 http://openmetrics.eu/openmetrics#Space_582
1 http://openmetrics.eu/openmetrics#RC04N0062_Oc... TUC.245.77.R104 http://openmetrics.eu/openmetrics#Space_787
2 http://openmetrics.eu/openmetrics#RC04N0063_Oc... TUC.245.77.R84 http://openmetrics.eu/openmetrics#Space_1317
3 http://openmetrics.eu/openmetrics#RC04N0064_Oc... TUC.245.77.R64 http://openmetrics.eu/openmetrics#Space_1147
4 http://openmetrics.eu/openmetrics#RC04N0065_Oc... TUC.245.77.R44 http://openmetrics.eu/openmetrics#Space_969

Restricting our search to just temperature sensors...

df = sparql_to_df(g, """SELECT ?sensor ?data_id ?location WHERE {
   ?sensor rdf:type/rdfs:subClassOf* brick:Temperature_Sensor .
   ?sensor ref:hasExternalReference/ref:hasTimeseriesId ?data_id .
   ?sensor brick:isPointOf ?location .
   ?location rdf:type/rdfs:subClassOf* brick:Location
}""")
df = get_data(df.data_id[:3], df.location[:3])
fig = px.line(df)
fig.update_layout(showlegend=False)

Get sensors and setpoints for the same space

df = sparql_to_df(g, """SELECT ?sensor_id ?sp_id ?location WHERE {
   ?sensor rdf:type/rdfs:subClassOf* brick:Temperature_Sensor .
   ?sensor ref:hasExternalReference/ref:hasTimeseriesId ?sensor_id .
   ?sensor brick:isPointOf ?location .
   
   ?sp rdf:type/rdfs:subClassOf* brick:Temperature_Setpoint .
   ?sp ref:hasExternalReference/ref:hasTimeseriesId ?sp_id .
   ?sp brick:isPointOf ?location .
   ?location rdf:type/rdfs:subClassOf* brick:Location
}""")
df.head()
sensor_id sp_id location
0 TUC.245.77.R80 TUC.245.77.R88 http://openmetrics.eu/openmetrics#Space_1317
1 TUC.245.77.R440 TUC.245.77.R448 http://openmetrics.eu/openmetrics#Space_1854
2 TUC.245.77.R20 TUC.245.77.R28 http://openmetrics.eu/openmetrics#Space_338
3 TUC.245.77.R380 TUC.245.77.R388 http://openmetrics.eu/openmetrics#Space_3457
4 TUC.245.77.R360 TUC.245.77.R368 http://openmetrics.eu/openmetrics#Space_3641

Plot sensor vs setpoint

space_name = 'http://openmetrics.eu/openmetrics#Space_3641'
# grab sensor+setpoint name
(sensor, sp, _) = df[df.location == space_name].values[0]
# get the data
d = get_data([sensor, sp], ['sensor', 'setpoint'])['2022-05-17':]
# plot it!
px.line(d, title=space_name)
# ax = d.plot(figsize=(15,5)) ; _ = ax.set_title(space_name)

Setpoint Deviation Analysis¶

Produce the distribution of deviation from setpoint for all spaces in the building

def avg_deviation(space_name):
    # get names of data points for the space
    (sensor, sp, _) = df[df.location == space_name].values[0]
    # get the actual data
    d = get_data([sensor, sp], ['sensor', 'setpoint'])
    # do a little cleaning
    d.dropna(inplace=True)
    # compute average absolute deviation
    return (d['setpoint'] - d['sensor']).abs().mean()

# now run for all spaces in our building
all_spaces = df.location.unique()
deviations = pd.DataFrame({'space': all_spaces,
                           'deviation': [avg_deviation(space) for space in all_spaces]})
deviations.sort_values(by='deviation', ascending=False).head(10)
space deviation
15 http://openmetrics.eu/openmetrics#Space_3110 5.800000
1 http://openmetrics.eu/openmetrics#Space_1854 5.005882
8 http://openmetrics.eu/openmetrics#Space_969 4.592857
9 http://openmetrics.eu/openmetrics#Space_1147 4.491667
5 http://openmetrics.eu/openmetrics#Space_3819 4.416667
16 http://openmetrics.eu/openmetrics#Space_2934 3.981818
6 http://openmetrics.eu/openmetrics#Space_582 3.655556
10 http://openmetrics.eu/openmetrics#Space_1680 3.600000
4 http://openmetrics.eu/openmetrics#Space_3641 3.341379
7 http://openmetrics.eu/openmetrics#Space_787 3.188889 
px.histogram(deviations, x='deviation', labels={'deviation': 'Degrees Celsius'})

Power Usage Intensity¶

Use BOT + Brick data to look at relationship between room area and how much power the things in the room contain:

# area of spaces
areas = sparql_to_df(g, """SELECT ?location ?area WHERE {
    ?location rdf:type/rdfs:subClassOf* brick:Space .
    ?location props:hasArea ?area
}""")
areas['area'] = areas['area'].astype(float)
areas.head()
location area
0 http://openmetrics.eu/openmetrics#Space_4137 2.971325
1 http://openmetrics.eu/openmetrics#Space_4300 2.536987
2 http://openmetrics.eu/openmetrics#Space_4468 3.801856
3 http://openmetrics.eu/openmetrics#Space_4636 4.528568
4 http://openmetrics.eu/openmetrics#Space_4793 2.440534 
# energy usage w/n the space
meters = sparql_to_df(g, """SELECT ?location ?point ?pointtype ?id WHERE {
    ?location rdf:type/rdfs:subClassOf* brick:Space .
    ?meter rdf:type/rdfs:subClassOf* brick:Electrical_Meter ;
           brick:hasLocation ?location ;
           brick:hasPoint ?point .
    ?point rdf:type ?pointtype ;
           ref:hasExternalReference/ref:hasTimeseriesId ?id .
}""")
meters.head()
location point pointtype id
0 http://openmetrics.eu/openmetrics#Space_582 http://openmetrics.eu/openmetrics#RC04N0061_Power https://brickschema.org/schema/Brick#Active_Po... TUC.245.77.R2
1 http://openmetrics.eu/openmetrics#Space_787 http://openmetrics.eu/openmetrics#RC04N0062_Power https://brickschema.org/schema/Brick#Active_Po... TUC.245.77.R102
2 http://openmetrics.eu/openmetrics#Space_1317 http://openmetrics.eu/openmetrics#RC04N0063_Power https://brickschema.org/schema/Brick#Active_Po... TUC.245.77.R82
3 http://openmetrics.eu/openmetrics#Space_1147 http://openmetrics.eu/openmetrics#RC04N0064_Power https://brickschema.org/schema/Brick#Active_Po... TUC.245.77.R62
4 http://openmetrics.eu/openmetrics#Space_969 http://openmetrics.eu/openmetrics#RC04N0065_Power https://brickschema.org/schema/Brick#Active_Po... TUC.245.77.R42 
# get all meter data (hourly max)
meter_data = get_data(meters['id'], meters['location'])\
                .resample('1H')\
                .max()\
                .melt(ignore_index=False)
meter_data.head()
location value
time
2022-05-14 16:00:00 http://openmetrics.eu/openmetrics#Space_582 0.0
2022-05-14 17:00:00 http://openmetrics.eu/openmetrics#Space_582 0.0
2022-05-14 18:00:00 http://openmetrics.eu/openmetrics#Space_582 0.0
2022-05-14 19:00:00 http://openmetrics.eu/openmetrics#Space_582 0.0
2022-05-14 20:00:00 http://openmetrics.eu/openmetrics#Space_582 0.0 
# join data, compute EUI
all_data = areas.merge(meters, on='location', how='inner')
all_data = all_data.merge(meter_data, on='location', how='right')
all_data['eui'] = all_data['area'] * all_data['value']
all_data['time'] = meter_data.index
all_data['location'] = all_data['location'].str.split('#').apply(lambda x: x[-1])

px.scatter(all_data, x='time', y='eui', color='location')